A5100443788- Privacy-Preserving Technologies in Data
- Stochastic Gradient Optimization Techniques
- Advanced Bandit Algorithms Research
- Cryptography and Data Security
- Mobile Crowdsensing and Crowdsourcing
- Sparse and Compressive Sensing Techniques
- Adversarial Robustness in Machine Learning
- Neural Networks and Applications
- Advanced Wireless Network Optimization
- Anomaly Detection Techniques and Applications
- Immune Response and Inflammation
- Internet Traffic Analysis and Secure E-voting
- Autoimmune Neurological Disorders and Treatments
- Maritime Transport Emissions and Efficiency
- Neuroscience and Neural Engineering
- Cryptographic Implementations and Security
- Advanced Clustering Algorithms Research
- Seismology and Earthquake Studies
- Recommender Systems and Techniques
- Epilepsy research and treatment
- Digital and Cyber Forensics
- Advanced Decision-Making Techniques
- Advanced Optimization Algorithms Research
- Machine Learning and Algorithms
- Structural Integrity and Reliability Analysis
Huazhong University of Science and Technology
2006-2024
PLA Information Engineering University
2022
Geological Exploration Institute of Shandong Zhengyuan
2022
University of California, Berkeley
2021
Yale University
2020
Tongji Hospital
2018
Union Hospital
2014-2015
Hunan University
2013
Wuhan University of Science and Technology
2010
Acute lung injury (ALI) is a severe illness with high rate of mortality. Maresin 1 (MaR1) was recently reported to regulate inflammatory responses. We used LPS-induced ALI model determine whether MaR1 can mitigate injury.Male BALB/c mice were injected, intratracheally, either LPS (3 mg·kg(-1) ) or normal saline (1.5 mL·kg(-1) ). After this, saline, low dose (0.1 ng per mouse) (1 given i.v. Lung evaluated by detecting arterial blood gas, pathohistological examination, pulmonary oedema, cell...
Acute lung injury (ALI) is characterized by inflammation and diffuse infiltration of neutrophils. Neutrophil apoptosis recognized as an important control point in the resolution inflammation. Maresin 1 (MaR1) a new docosahexaenoic acid-derived proresolving agent that promotes However, its function neutrophil unknown. In this study, isolated human neutrophils were incubated with MaR1, pan-caspase inhibitor z-VAD-fmk, lipopolysaccharide (LPS) to determine mechanism apoptosis. was induced...
Resistance is one of the important performance indicators ships. In this paper, a prediction method based on Radial Basis Function neural network (RBFNN) proposed to predict resistance 13500 transmission extension unit (13500TEU) container ship at different drafts. The predicted draft state in known range called interpolation prediction; otherwise, it extrapolation prediction. First, features are extracted make Rt results show that RBFNN significantly better than other four machine learning...
With the advent of era big data, deep learning has become a prevalent building block in variety machine or data mining tasks, such as signal processing, network modeling and traffic analysis, to name few. The massive user crowdsourced plays crucial role success models. However, it been shown that may be inferred from trained neural models thereby exposed potential adversaries, which raises information security privacy concerns. To address this issue, recent studies leverage technique...
Online optimization has been a successful framework for solving large-scale problems under computational constraints and partial information. Current methods online convex require either projection or exact gradient computation at each step, both of which can be prohibitively expensive applications. At the same time, there is growing trend non-convex in machine learning community need methods. Continuous DR-submodular functions, exhibit natural diminishing returns condition, have recently...
Federated learning (FL) has attracted increasing attention in recent years due to its data privacy preservation and great applicability large-scale user scenarios. However, when FL faces numerous clients, it is inevitable emerge the non-independent identically distributed (non-iid) between which brings an enormous challenge for model training performance analysis like convergence. Besides, non-iid data, participating clients of tend be extremely heterogeneous so number samplings among causes...
Deep learning (DL) has been adopted in a broad range of Internet-of-Things (IoT) applications such as auto-driving, intelligent healthcare and smart grids, but limitations those relating to data user privacy can complicate its broader implementation. Seeking jointly address both utility, this paper we connect the layer-wise relevance propagation with gradient descent for injecting proper noise into gradients. We also improve conventional clipping method by dividing gradients several groups;...
This paper explores the generalization loss of linear regression in variably parameterized families models, both under-parameterized and over-parameterized. We show that curve can have an arbitrary number peaks, moreover, locations those peaks be explicitly controlled. Our results highlight fact classical U-shaped recently observed double descent are not intrinsic properties model family. Instead, their emergence is due to interaction between data inductive biases learning algorithms.
As an important part of computer forensics, network forensics particularly places emphasis on dynamic information collection and proactive defense. Most systems based intrusion detection or honeypot rarely emphasize the availability actual servers. In addition, few them discussed occasion particularly. The work presented in this paper is idea to assist with tolerance deception technology enhance server system gather more useful evidences a proper occasion. A mechanism proposed modeled finite...
This paper proposes a connection weighting scheme of complex-valued Hopfield neural network for associative memory constrained by given attractive domain.Both equilibrium conditions and stability analysis results are used in the synthesis procedure.We solve equation singular value decomposition technique obtain general solution weight matrix with free sub-matrix.Such parameter corresponding to domain contained inequations which derived from can be represented as linear (LMIs).The such LMIs...
This paper considers online convex optimization (OCO) with generated i.i.d. stochastic constraints, where the performance is measured by adaptive regret. The constraints are disclosed at each round to learner after decision made. Different from previous non-adaptive constrained OCO algorithm which directly generalized static gradient descent algorithm, we propose novel Virtual Queue-based Following-the-Leader-History (VFLH) strategy make adaptive. In this framework, generalizes experts that...
Due to the data stream is real-time, fast, unlimited, one-pass, clustering requires algorithms which are capable process in limited time and memory. In this paper, we propose a algorithm based on improved similarity search tree (SSMC-Tree), introduce buffer, hitchhike processing local aggregation strategy, it can adapt different speed stream. We adopt an outlier mechanism by introducing potential core-micro-cluster buffer micro-cluster noise Experimental results show that our high-speed with noise.
Abstract Microseismic event detection helps to predict outbreak catastrophic problems and has essential applications in resource exploration. Low SNR microseismic signal is a challenging task detection. In this paper, we propose (convolutional neural network method based on variance fractal dimension) VFD-CNN the dimension (VFD). method, signals background noise are first measured by dimension, which can effectively extract seismic nonlinear features. These features then fed into distinguish...
As numerous consumer electronics applications like smartphones and wearables generate lots of distributed data daily, desire to safely efficiently tackle private isolated data. Federated learning (FL) is hopeful satisfy the above requirement due strong security applicability large-scale scenarios. But diverse clients inevitably cause non-independent identically (non-iid) among clients, which severely hinders performance analysis. Besides, affected by non-iid data, participating are typically...
The neuronal network cultured in virto used as a important tool for brain study have been realized by more and people owing to its non-invasive nature. But till now, there isn't parameter that conveniently describes the changing states of from whole. In this paper, synchrony calculation acted reactive results electrical stimulation (used learning training) or bicuculine is analyzed variety tried value depict diversification network. These experimental processed way were given out end paper.
Federated learning (FL) is a promising distributed framework for collaborative artificial intelligence model training while protecting user privacy. A bootstrapping component that has attracted significant research attention the design of incentive mechanism to stimulate collaboration in FL. The majority works adopt broker-centric approach help central operator attract participants and further obtain well-trained model. Few consider forging participant-centric among pursue an FL their common...
Much of the value that IoT (Internet-of-Things) devices bring to ``smart'' homes lies in their ability automatically trigger other devices' actions: for example, a smart camera triggering lock unlock door. Manually setting up these rules or applications, however, is time-consuming and inefficient. Rule recommendation systems can suggest users by learning which are popular based on those previously deployed (e.g., others' homes). Conventional formulations require central server record used...
Deep learning techniques have been widely used in the field of Side Channel Attack (SCA), which poses a serious threat to security cryptographic algorithms. However, deep learning-based side channel attack also has problems such as inefficient models, poor robustness, and longtime consumption. To address these problems, this paper focuses on performance Long Short-term Memory(LSTM) combining with dimensional compression technique Sparse Auto Encoder (SAE), validates it fully synchronized...